2.4 Primary outcome: reduction in anhedonia scores at 8 weeks (from 4 to 12 weeks)
2.4.1 Pairwise meta-analysis
Figure 2. Forest plot for symptoms of anhedonia (primary outcome) comparing pro-dopaminergic interventions vs placebo for individuals with anhedonia at 4-12 weeks (primary timepoint). SMD: standardised mean difference, 95% CI: 95% confidence intervals, SD: standard deviation.
The effect of pro-dopaminergic drugs versus placebo showed an effect favouring pro-dopaminergic interventions with a SMD of -0.244 (95% CI from -0.456 to -0.031). There is some heterogeneity as shown by the 95% prediction interval from -0.74 to 0.252.
2.4.2 Risk of bias
Figure 3. Risk of bias assessment.
Evidence for the efficacy of pro-dopaminergic interventions vs placebo were rated as having a range in their overall risk of bias. Two studies (33%) were assessed as having ‘high’ risk of bias due to having ‘high’ risk of bias in the missing outcome data domain and selection of the reported results domain. Two studies (33%) had an overall ‘moderate’ risk of bias rating as they had ‘some concerns’ in two domain ratings. The remaining two studies were rated as having a ‘low’ overall risk of bias.
2.4.3 Reporting bias
The extent to which the result was affected by reporting biases was rated as low as per the RoB-ME assessment (Page et al. 2023). This was as the potential for missing studies across the review was judged to be low. In addition, none of the included studies were deemed to have generated an eligible result that was not reported, and no studies were judged to be unclear as to whether they generated an eligible result that was not reported. We made this decision based on the results matrix we generated in step 2 of the RoB-ME tool (Page et al. 2023).
2.4.4 Meta-regression analyses
We did not perform any meta-regressions as the total number of studies was below 10.
2.4.5 Post-hoc analyses
We had access to individual participant data (IPD) of randomised controlled trials on depression from a NIHR-funded project. We performed a series of post-hoc analyses on the MADRS “inability to feel” item (aggregated IPD from 34 studies, 14054 participants). First, we included all the available studies and compared all the antidepressants against placebo.
### 2.4.5.1 Pro-dopaminergic interventions and non-dopaminergic drugs vs placebo, random effects meta-analysis
Figure 4. Forest plot for symptoms of anhedonia (primary outcome) comparing bupropion versus placebo for individuals with anhedonia at 4-12 weeks (primary timepoint). SMD: standardised mean difference, 95% CI: 95% confidence intervals, SD: standard deviation.
Figure 5. Comparative effect sizes of antidepressants versus placebo.
The mean effect of bupropion (pro-dopaminergic intervention) versus placebo was -0.12 (SMD, 95% CI from -0.25 to 0.00; 34 studies, 14054 participants). In comparison, we found effects across non-dopaminergic interventions to range from -0.09 (SMD, 95% CI from -0.78 to 0.60) in amitryptiline to -0.40 (SMD, 95% CI from -0.48 to -0.32) in duloxetine.
2.4.5.2 Aggregated IPD, bupropion versus placebo, random effects pairwise meta-analysis
We also performed a random effects pairwise meta-analysis of bupropion versus placebo based on the aggregated IPD of the ‘inability to feel’ item on the MADRS scale, resulting in estimates comparable to the network meta-analytical model (SMD -0.12, 95% CI from -0.29 to 0.05; 4 studies, 1085 participants).
2.4.5.3 Aggregated IPD and early studies identified in this living systematic review, random effects pairwise meta-analysis
Finally, we performed a random effects pairwise meta-analysis including aggregate data from both sources (IPD and retrieved as aggregated). As three studies (Hewett 2009 - 87997883, Hewett 2010a - 87997755, Koshino 2013 - 87997374) were available in both sources, we prioritised aggregate IPD over data retrieved as aggregated. The mean effect of bupropion versus placebo was -0.19 (SMD, 95% CI from -0.33 to 0.04; 6 studies, 2489 participants)
